I live in the worst area in the UK for dating

Analyzing and plotting the 2021 Census data instead of, you know, going outside.

Introduction

Where are all the women at?

I live in Cambridge and it's okay. There are infrastructure and housing problems but where aren't there these days? It's a university city too. Sure, they only started giving degrees to women in 1948 but, 76 years on, you'd think the gender ratio would have balanced out. And anyway, undergraduate students aren't in my dateable range any more and the population that is isn't necessarily just the student population but older.

But I started having this suspicion. Things were just a bit too off. On the streets, on dating apps, in grocery stores. I had to know.

The government had to know too. That's why they do a Census every ten years, with the most recent one being in 2021. It collects plenty of data, not just sex but also demographics, occupation, housing, travel, education, health etc, with over a hundred variables, broken down by local authority, middle layer super output areas and output areas (basically, a block).

But the built-in visualizers, like the one on the Office for National Statistics' website, don't allow you to visualize more than one variable or condition the value on another variable. So, I could only get the gender ratio in my local authority and not the gender ratio in a certain age range. And that 49.9% female-to-male ratio that I saw in the data browser sounded too clean. Too convenient.

It was time to download the data and figure it out myself. How bad was the city of Cambridge really in terms of the gender ratio?

I had no idea.

Acquiring the data

Step one, go to the ONS's website and get the TS009 dataset. There's also a more granular version, RM200, that goes down to the individual output area, but that's way too granular (too noisy and if need be, I can always travel around beyond my street).

import pandas as pd
import seaborn as sns
sns.set_theme()
import geopandas as gpd
import matplotlib
from matplotlib import pyplot as plt
%matplotlib inline

csv = pd.read_csv("TS009-2021-2.csv")
csv.columns = ["la_code", "la", "sex_code", "sex", "age_code", "age", "observation"]

Calculating the stats

Step two, draw the rest of the owl.

FEMALE_AGE_RANGE = (25, 30)
MALE_AGE_RANGE = (25, 30)

We're interested in women aged 25-30. We're also interested in our competition, men aged 25-30 (well, that might be different for men than for women but let's do it this way in the name of symmetry).

This obviously has a bunch of other assumptions: that all these people are single or at least equally single and equally monogamous and/or equally homosexual. But it's a start.

Now, let's get our counts and the gender ratio:

def get_stats(grp):
    return pd.Series(
        {
            f"females_{FEMALE_AGE_RANGE[0]}_to_{FEMALE_AGE_RANGE[1]}": grp[
                (
                    (grp["age_code"] >= FEMALE_AGE_RANGE[0])
                    & (grp["age_code"] <= FEMALE_AGE_RANGE[1])
                )
                & (grp["sex"] == "Female")
            ]["observation"].sum(),
            f"males_{MALE_AGE_RANGE[0]}_to_{MALE_AGE_RANGE[1]}": grp[
                (grp["age_code"] >= MALE_AGE_RANGE[0])
                & (grp["age_code"] <= MALE_AGE_RANGE[1])
                & (grp["sex"] == "Male")
            ]["observation"].sum(),
            "total_population": grp["observation"].sum(),
        }
    )


stats = csv.groupby(["la_code", "la"]).apply(get_stats).reset_index()
stats["female_to_male_ratio"] = (
    stats[f"females_{FEMALE_AGE_RANGE[0]}_to_{FEMALE_AGE_RANGE[1]}"]
    / stats[f"males_{MALE_AGE_RANGE[0]}_to_{MALE_AGE_RANGE[1]}"]
)
stats = stats.sort_values("female_to_male_ratio")

Plotting the stats

So, let's plot the top local authorities by gender ratio.

ax = sns.barplot(stats[:-21:-1], y="la", x="female_to_male_ratio")
ax.bar_label(ax.containers[0], fmt="%.2f")

Sounds reasonable. A lot of boroughs in London as well as the Isles of Scilly (population: 2,100).

Now, let's plot the bottom 20 LAs by gender ratio.

ax = sns.barplot(stats[:20], y="la", x="female_to_male_ratio")
ax.bar_label(ax.containers[0], fmt="%.2f")

Cool. Now let's find Cambridge... oh.

Uh oh.

Turns out, there are roughly 9 women aged 25-30 for 10 men aged 25-30 in Cambridge. But it gets worse. If we take the number of men aged 25-30 and subtract it from the number of women, we get this:

stats["surplus"] = (
    stats[f"females_{FEMALE_AGE_RANGE[0]}_to_{FEMALE_AGE_RANGE[1]}"]
    - stats[f"males_{MALE_AGE_RANGE[0]}_to_{MALE_AGE_RANGE[1]}"]
)
ax = sns.barplot(stats.sort_values("surplus")[:20], y="la", x="surplus")
ax.bar_label(ax.containers[0], fmt="%.2f")
plt.xlim(-1000, 0)
plt.title("Bottom 10 local authorities by excess women - men")

Yep, Cambridge is the worst area in the UK for dating, assuming you want to date women aged 25-30.

Plotting the data on a map

Let's plot this on a map, using this blog post as inspiration.

First, get the 2021 local authority boundaries from the Open Geography Portal. We're interested in the Shapefile format.

lad_geom = gpd.read_file("LAD_DEC_2021_GB_BGC.shp")
stats_geom = lad_geom.merge(stats, right_on="la_code", left_on=["LAD21CD"])

Make a nice Choropleth map of the gender ratios:

from matplotlib import pyplot as plt
from mpl_toolkits.axes_grid1 import make_axes_locatable
import matplotlib.colors as colors

fig = plt.figure(figsize=(10, 15))
plt.title(
    f"England, Local Authorities by gender ratio of people aged {FEMALE_AGE_RANGE[0]}-{FEMALE_AGE_RANGE[1]}"
)
ax = fig.gca()
ax.set_axis_off()
divider = make_axes_locatable(ax)
cax = divider.append_axes("right", size="5%", pad=0.0)
stats_geom[["female_to_male_ratio", "geometry"]].plot(
    column="female_to_male_ratio",
    linewidth=0,
    ax=ax,
    cax=cax,
    legend=True,
    norm=colors.CenteredNorm(vcenter=1.0),
    cmap="coolwarm",
)

London is king here, just like we saw in the previous visualization.

Now, the map with the absolute difference between the number of women and men:

fig = plt.figure(figsize=(10, 15))
plt.title(
    f"England, Local Authorities by number of women {FEMALE_AGE_RANGE[0]}-{FEMALE_AGE_RANGE[1]} - men {MALE_AGE_RANGE[0]}-{MALE_AGE_RANGE[1]} (excess)"
)
ax = fig.gca()
ax.set_axis_off()
divider = make_axes_locatable(ax)
cax = divider.append_axes("right", size="5%", pad=0.0)

stats_geom["surplus"] = (
    stats[f"females_{FEMALE_AGE_RANGE[0]}_to_{FEMALE_AGE_RANGE[1]}"]
    - stats[f"males_{MALE_AGE_RANGE[0]}_to_{MALE_AGE_RANGE[1]}"]
)

stats_geom[["surplus", "geometry"]].plot(
    column="surplus",
    linewidth=0,
    ax=ax,
    cax=cax,
    legend=True,
    cmap="Reds",
)

Not much is surprising here. The one red hot spot in the West Midlands is Birmingham.

Full table

Want to see how your local authority ranks? See the full table under the foldout:

Conclusion

There's a bunch of other data that's collected by the Census and existing visualizers are mostly about visualizing single variables on a map. Perhaps it would be cool to build something like what I did, but interactive, with filters and dynamic expressions. Or a different visualization method, like the demographic pyramid in a certain area.

But maybe later. I suddenly feel a weird urge to go to the gym or move out.